Cellular/Molecular

The Roles of Cdk5-Mediated Subcellular Localization ofFOXO1 in Neuronal Death

X Jiechao Zhou,1* Huifang Li,1* Xiaoping Li,1 Guanyun Zhang,1 Yaqiong Niu,1 Zengqiang Yuan,2 Karl Herrup,3,4

Yun-Wu Zhang,1 Guojun Bu,1 Huaxi Xu,1 and Jie Zhang1

1Fujian Provincial Key Laboratory of Neurodegenerative Disease and Aging Research, Institute of Neuroscience, Medical College, Xiamen University,Xiamen, 361005, Fujian Province, China, 2Institute of Biophysics, Chinese Academy of Sciences, Chaoyang District, Beijing, 100101, China, 3Division of LifeScience and the State Key Laboratory of Molecular Neuroscience, Hong Kong University of Science and Technology, Kowloon, Hong Kong, and4Department of Cell Biology and Neuroscience, Rutgers University, Piscataway, New Jersey 08854

Deficiency of cyclin-dependent kinase 5 (Cdk5) has been linked to the death of postmitotic cortical neurons during brain development.We now report that, in mouse cortical neurons, Cdk5 is capable of phosphorylating the transcription factor FOXO1 at Ser249 in vitro andin vivo. Cellular stresses resulting from extracellular stimulation by H2O2 or �-amyloid promote hyperactivation of Cdk5, FOXO1 nuclearexport and inhibition of its downstream transcriptional activity. In contrast, a loss of Cdk5 leads to FOXO1 translocation into the nucleus:a shift due to decreased AKT activity but independent of S249 phosphorylation. Nuclear FOXO1 upregulates transcription of the proapo-ptotic gene, BIM, leading to neuronal death, which can be rescued when endogenous FOXO1 was replaced by the cytoplasmically localizedform of FOXO1, FOXO1-S249D. Cytoplasmic, but not nuclear, Cdk5 attenuates neuronal death by inhibiting FOXO1 transcriptionalactivity and BIM expression. Together, our findings suggest that Cdk5 plays a novel and unexpected role in the degeneration of postmi-totic neurons through modulation of the cellular location of FOXO1, which constitutes an alternative pathway through which Cdk5deficiency leads to neuronal death.

Key words: Cdk5; FOXO1; neuronal death; phosphorylation

IntroductionCyclin-dependent kinase 5 (Cdk5) is a unique member of theCdk family. Although its cloning was based on its sequence ho-mology to other Cdks (Hellmichet al., 1992), Cdk5 is not acti-vated by traditional cyclins but by the cyclin paralogs p35 and p39(Lew et al., 1994; Tsai et al., 1994; Dhavan and Tsai, 2001). Cdk5is ubiquitously expressed, yet the high neuronal expression of p35and p39 largely confines physiological Cdk5 enzyme activity to alimited number of tissues, most prominently the nervous system.Cdk5 is expressed early in embryo genesis and plays pivotal rolesin neuronal development, including survival, migration, and

maturation (Ohshima et al., 1996, 1999; Gilmore and Herrup,2001; Cruz and Tsai, 2004; Hawasli and Bibb, 2007; Hisanaga andEndo, 2010).

In addition to its developmental functions, in the mature nervoussystem, Cdk5 is involved in synaptic function and neuronal survival.For example, hyperactive Cdk5 is reported to promote neuronaldeath and has been proposed to play a role in the pathogenesis ofAlzheimer’s disease (AD). In AD, the overactivation of Cdk5 isthought to induce tau phosphorylation and tangle formation, whichpromotes neuron loss. One mechanism by which the hyperactiva-tion of Cdk5 might occur is through the proteolytic cleavage of p35by the calcium-dependent protease calpain to create p25. This cleav-age is prominent in the neuronal response to �-amyloid-inducedstress (Patrick et al., 1999; Su and Tsai, 2011). Although too muchCdk5 activity might be bad, too little Cdk5 is also dangerous to thehealth and survival of adult neurons. For example, there is floridneuronal death in Cdk5�/� mouse cortex (Cicero and Herrup,2005). Cdk5-deficient neurons are also more sensitive to extracellu-lar stressors, such as UV irradiation (Li et al., 2002) and �-amyloidchallenge. In forebrain, conditional Cdk5 knock-out leads to a pro-gressive neuronal degeneration in adult brain (Takahashi et al.,2010). Restoring Cdk5 function in Cdk5�/� neurons reverses theirchromatolytic morphology (Tanaka et al., 2001; Cheung and Ip,2004). To date, however, the mechanism by which the loss of Cdk5leads to neuronal death remains unclear. Although cell cycle regula-tion undoubtedly plays a part, there are suggestions that indepen-dent pathways are also at work.

Received July 15, 2014; revised Dec. 29, 2014; accepted Dec. 31, 2014.Author contributions: J. Zhou, K.H., and J. Zhang designed research; J. Zhou, H.L., X.L., G.Z., Y.N., and J. Zhang

performed research; Z.Y. and Y.-W.Z. contributed unpublished reagents/analytic tools; J. Zhou, H.L., K.H., G.B., H.X.,and J. Zhang analyzed data; K.H., G.B., H.X., and J. Zhang wrote the paper.

This work was supported by Xiamen University 985 Project, National Science Foundation of China Grant81271421 to J. Zhang and Grants 91332114 and U1405222 to H.X., Natural Science Foundation of Fujian Province ofChina Grant 2013J01147 and Grant 2014J06019 to J. Zhang, Program for New Century Excellent Talents in University,Hong Kong University of Science and Technology, National Key Basic Research Program of China 2013CB530900, andResearch Grants Council HKSAR HKUST12/CRF/13G and GRF660813.

The authors declare no competing financial interests.*J. Zhou and H.L. contributed equally to this work as joint first authors.Correspondence should be addressed to either of the following: Dr. Jie Zhang, Fujian Provincial Key Laboratory of

Neurodegenerative Disease and Aging Research, Institute of Neuroscience, Medical College, Xiamen University,Xiamen, 361005, China, E-mail: jiezhang@xmu.edu.cn; or Dr. Karl Herrup, Division of Life Science and the State KeyLaboratory of Molecular Neuroscience, Hong Kong University of Science and Technology, Kowloon, Hong Kong.E-mail: herrup@ust.hk.

DOI:10.1523/JNEUROSCI.3051-14.2015Copyright © 2015 the authors 0270-6474/15/352624-12$15.00/0

2624 • The Journal of Neuroscience, February 11, 2015 • 35(6):2624 –2635

The FOXO (forkhead) transcription factors have a broad im-pact on various cellular processes, including cell cycle regulation,apoptosis, DNA repair, and autophagy (Lehtinen et al., 2006;Yuan et al., 2009; Xie et al., 2012). Activation of FOXO promotescell death by regulating the expression of genes involved in apo-ptosis, such as BIM, FAS ligand, and TRAIL (Modur et al., 2002;Essafi et al., 2005; Real et al., 2005). In the current study, we reportthat FOXO1 is a key link in a cell cycle-independent chain ofevents leading from Cdk5 deficiency to the death of CNS neu-rons. Our data demonstrate that, in Cdk5-deficient postmitoticneurons, FOXO1 phosphorylation at Ser249 is lost. This initiatesthe expression of proapoptotic genes and facilitates neuronaldeath.

Materials and MethodsAntibodies and chemical reagents. Antibodies against FOXO1 and cleavedcaspase-3 were from Cell Signaling Technology. Anti-Cdk5, GFP, p35,hnRNP A2/B1, 14-3-3�, and GAPDH antibodies were from Santa CruzBiotechnology. Anti-�-tubulin and anti-Flag (M2) were from Sigma-Aldrich. Anti-BrdU and FOXO1 antibodies were from Abcam. Phospho-FOXO1-S319, phospho-AKT (Ser473), and Akt1 were from SangonBiotech. The phospho-FOXO1-S249 antibody was described previously(Yuan et al., 2008). Secondary antibodies used for immunocytochemis-try were as follows: goat anti-mouse Alexa-488 and -594 and goat anti-rabbit Alexa-488 and -594 (Invitrogen).

DAPI was used as a nuclear counterstain at 1 mg/ml; LY-294002 hy-drochloride, BrdU, and roscovitine were purchased from Sigma-Aldrich.The dual-luciferase reporter assay kit was purchased from Promega.

Animals. A colony of Cdk5�/� mice was maintained on a mixedC57BL/6J background. Homozygous mutant embryos of both sexes wereproduced by intercrossing heterozygous Cdk5�/� mice. Wild-typeC57BL/6J mice were purchased from SLRC Laboratory. Timed pregnan-cies were established, and the embryos were taken at embryonic day 16.5for primary cortical cultures or brain lysates. We assume each litter con-tained a nearly equal mixture of the two sexes, although no attempt wasmade to assess the sex of the harvested embryos. All animal procedureswere performed in accordance with Xiamen University Institutional An-imal Care and Use Committee standards.

Cell culture. Neuro2A and 293T cells were cultured in DMEM mediumcontaining 10% FBS, 100 U/ml penicillin, and 100 �g/ml streptomycin.Cells were maintained at 37°C and 5% CO2. Primary cortical neuronswere cultured from cortex of E16.5 d mice. For Cdk5-deficient cultures,all of the embryos from a Cdk5 �/�xCdk5 �/� mating were harvested andtreated separately as described previously (Zhang et al., 2010a). Primarycortical neurons were cultured in Neurobasal medium supplementedwith B27 and 2 mM glutamine and grown on coverslips or plastic cultureplate coated with poly-L-lysine (25 �g/ml).

Plasmids and small interference RNA (siRNA). The generosity of manylaboratories allowed us to assemble a majority of the vectors used in thisstudy. Pcmv-Flag-FOXO1-AAA and sh-foxo vectors were kindly pro-vided by Dr. You Han (Xiamen University, Xiamen); Bim-luciferase and3xIRS-luciferase reporter genes vectors were provided by Dr. KL Guan(University of California, San Diego). GST-FOXO1 was provided by Dr.Haojie Huang (Mayo Clinic, Rochester, MN). CMV-p35 was a gift fromDr. Li-Huei Tsai (Massachusetts Institute of Technology, Cambridge,MA). Flag-FOXO1 was purchased from Addgene. GFP-Cdk1 to GFP-Cdk6, GFP-cyclinA, GFP-cyclinB, GFP-cyclinC, and GFP-cyclinD were

Figure 1. Cdk5 phosphorylates FOXO1 at serine 249. A, B, Lysates of 293T cells transfected with Cdk5/p35, Cdk5/p25, Cdk5-KD/p35, Cdk5-KD/p25 (Cdk5 kinase-dead plasmid), Cdk1/Cyclin B,Cdk2/Cyclin A, or control vector, together with GFP-FOXO1 or the GFP-FOXO1–S249A mutant, were analyzed by immunoblotting with antibodies to pS249-FOXO1, FOXO1, Cdk5, p35/p25, GFP, orGAPGH as shown. C, D, In vitro kinase assays with immunoprecipitated (IP) Cdk5. Recombinant GST-FOXO1 or GST-FOXO1–S249A proteins were subjected to an in vitro kinase assay with IP GFP fusionproteins from 293T cells expressing the indicated constructs. Kinase reaction products were immunoblotted with the pS249-FOXO1 antibody. FOXO1 proteins levels were evaluated by CoomassieBlue staining. E, The interactions among FOXO1, Cdk5, and p35 were investigated in 293T cells. The 293T cells were cotransfected Flag-FOXO1 with GFP-Cdk5 and/or GFP-p35. The antibodies usedfor IP are indicated in the top of the gel. After immunoprecipitation, the gels were blotted with indicated antibodies. F, Primary cortical neurons were transfected with the GFP-FOXO1-WT orGFP-FOXO1–S249A plasmid together with p25/Cdk5 or p25/Cdk5-KD or control vector in combination with the 3xIRS luciferase reporter system and subjected to luciferase assay. Data are mean �SEM; n � 5. *p � 0.05. G, Primary cortical neurons were transfected with different Cdk/cyclin pairs together with the 3xIRS-luciferase reporter gene and subjected to luciferase assays. Data aremean � SEM; n � 5. *p � 0.05.

Zhou, Li et al. • Cdk5 Mediates FOXO1 Localization J. Neurosci., February 11, 2015 • 35(6):2624 –2635 • 2625

Figure 2. Oxidative stress induces FOXO1 nuclear export by active Cdk5. A, Lysates of 293T cells transfected with GFP-FOXO1 or GFP-FOXO1–S249A, together with Cdk5/p25 or thecontrol vector, were immunoprecipitated with the FOXO1 antibody and immunoblotted with the FOXO1, Cdk5, and 14-3-3� antibodies. B, Primary cortical neurons were treated with 300�M H2O2 for the indicated time. Whole-cell lysates were then analyzed by immunoblotting with the FOXO1, pS249-FOXO1, p35/p25, and �-tubulin antibodies. C, Quantifications of therelative pS249-FOXO1 intensities. Data are mean � SEM; n � 3. *p � 0.05. D, E, Primary cortical neurons were treated with 300 �M H2O2 for the indicated time. The cytosolic and nuclearextracts were isolated and immunoblotted with the indicated antibodies. E, Quantifications of the relative nuclear FOXO1 intensities. Data are mean � SEM; n � 3. *p � 0.05. F, Primarycortical neurons were treated with 50 �M roscovitine or vehicle control (DMSO). Endogenous FOXO1 and Map2 (Microtubule-associated protein 2) were immunostained to investigate thelocalization of FOXO1 in neurons (Map2-positive cells). G–I, GFP-FOXO1 or GFP-FOXO1–S249D was transfected in primary cortical neurons. G, After 50 �M roscovitine administration, thenuclear translocation time of FOXO1 proteins (GFP-tag) was monitored by live cell imaging. Representative images and quantifications of subcellular localization of GFP-FOXO1 andGFP-FOXO1–S249D are shown in H and I, respectively. J–M, Primary cortical neurons were transfected with control U6 or shCdk5 plasmid together with GFP-FOXO1-WT or GFP-FOXO1–S249A in combination with the 3xIRS luciferase reporter system. On DIV 7, the neurons were treated with 300 �M H2O2 for 3 h or 3 �M �-amyloid for 10 h, and then subjected to luciferaseassay. Data are mean � SEM; n � 5. *p � 0.05.

2626 • J. Neurosci., February 11, 2015 • 35(6):2624 –2635 Zhou, Li et al. • Cdk5 Mediates FOXO1 Localization

generated by inserting target genes into pEGFP-C3 vector (Clontech). GFP-FOXO1-S249A, GFP-FOXO1-S249D, GFP-FOXO1-AAA-S249A, GFP-FOXO1-AAA-S298A, GFP-FOXO1-AAA-S249D, and Flag-tagged FOXO1mutations were generated by site-directed mutagenesis (Stratagene).CMV-mCherry-IRES-U6-scramble and CMV-mCherry-IRES-U6-shcdk5were from GENECHEM. The Flag-14-3-3� plasmid was cloned by PCRinto the pFLAG-CMV-2 expression vector. GFP, GFP-p25, GFP-p35,GFP-Cdk5, GFP-Cdk5-KD (D144N), GFP-NLS, GFP-Cdk5-NLS, GFP-NES, and GFP-Cdk5-NES vectors were described previously (Zhang etal., 2010a, 2012). Gene-specific siRNAs targeting Cdk5 and foxo1 werepurchased from Ribobio.

Luciferase assay. Cells were transfected with 1 �g of 3xIRS or Bim-luciferase reporter plasmid and 200 ng of the tk-renilla luciferase reporterplasmid, the latter serving as an internal control. At 72 h after transfec-tion, cells were left untreated or treated with drugs. The cells were thenwashed with PBS and lysed in the buffer provided in the Promega lu-ciferase system and subjected to luciferase assays. Relative luciferase ac-tivities were obtained by normalizing the firefly luciferase activity againstrenilla luciferase activity.

Western blotting and coimmunoprecipitation. Immunoblotting and co-immunoprecipitation were performed as described previously (Zhang etal., 2010a). Briefly, harvested cells were homogenized in ice-cold NP-40lysis buffer with protease inhibitor mix. The samples were centrifuged at12,000 � g for 15 min at 4°C. The supernatant was collected, and totalprotein levels were measured by the micro BCA protein assay kit(Thermo Fisher Scientific). Fractionation of cells into cytoplasmic andnuclear components was accomplished with an NER-mammalian kitfrom Thermo Fisher Scientific according to the manufacturer’s instruc-tions. For Western blots, the lysates were separated with SDS-PAGE andelectrophoretically transferred onto nitrocellulose membranes. Themembranes were blocked with 5% nonfat milk in TBST and probed withprimary antibodies, followed by treatment with HRP-linked secondaryantibodies and ECL Western blotting detection reagents. The intensity ofimmunoreactive bands was quantified using National Institutes ofHealth ImageJ software. For immunoprecipitation, the cell lysates were

incubated with immunoprecipitation antibody at 4°C for 4 h, followedby overnight incubation with protein G. The beads were washed fivetimes with ice-cold NP-40 lysis buffer, and the bound proteins wereanalyzed by SDS-PAGE and immunoblotting analysis.

Immunofluorescence staining. The neuronal cells were immunostainedas described previously (Zhang et al., 2010a). In brief, the cells were fixedwith 4% PFA for 20 min. After three washes with PBS, the cells wereblocked with 10% goat serum in PBS containing 0.5% Tween 20 toreduce nonspecific antibody binding. Neurons were then incubated withthe primary antibody at 4°C overnight. After washing with PBS fourtimes, AlexaFluor-488- or 546-conjugated secondary antibody was usedto detect the signal. Nuclear morphology was visualized using the DNAdye DAPI.

Statistical analysis. All data were obtained from at least three differentpreparations. Statistical analysis of the data was performed with a two-tailed Student’s t test or a one-way ANOVA using Graphpad Prism 5.Data are presented as the mean � SEM. p � 0.05 was taken to representstatistical significance.

ResultsCdk5 associates with FOXO1 and phosphorylates serine 249We coexpressed Cdk5 with p35 or its cleaved form p25 in 293Tcells together with wild-type human FOXO1 or a nonphosphor-ylatable human FOXO1-S249A mutant. Using a previously gen-erated antibody (Yuan et al., 2008) that specifically recognizesFOXO1 phosphorylated at Ser249, we immunoblotted gels oftotal cell lysates. We found that both Cdk5/p35 and Cdk5/p25transfection can dramatically increase the amount of phosphor-S249-FOXO1 labeling. No labeling was detected when the mu-tant FOXO1-S249A was used, even in the presence of Cdk5/p35or Cdk5/p25 cotransfection (Fig. 1A). The kinase dead-mutationof Cdk5 (Cdk5-KD) complex, such as Cdk5-KD/p35 or Cdk5-KD/p25, was unable to phosphorylate FOXO1 at Ser249 (Fig.

Figure 3. Ser249-FOXO1 is less affected by AKT inhibition. A–F, GFP-FOXO1, GFP-FOXO1–S249A, or GFP-FOXO1–S249D was transfected in primary cortical neurons treated with DMSO orLY294002 for 16 h. Representative images and the quantification of subcellular localization of GFP-FOXO1, GFP-FOXO1–S249A, and GFP-FOXO1–S249D were shown in top and bottom panels,respectively. G–I, GFP-FOXO1-AAA, GFP-FOXO1-AAA-S249A, or GFP-FOXO1-AAA-S249D was transfected in primary cortical neurons. Representative images and the quantification of subcellularlocalization of GFP-FOXO1-AAA, GFP-FOXO1-AAA-S249A, and GFP-FOXO1-AAA-S249D were shown in top and bottom panels, respectively. J–K, Primary cortical neurons were transfected withGFP-FOXO1-AAA, GFP-FOXO1-AAA-S249A, or GFP-FOXO1-AAA-S249D plasmids together with the 3xIRS-luciferase (J ) or Bim-luciferase (K ) reporter gene and subjected to luciferase assays. Data aremean � SEM; n � 5. *p � 0.05.

Zhou, Li et al. • Cdk5 Mediates FOXO1 Localization J. Neurosci., February 11, 2015 • 35(6):2624 –2635 • 2627

1A). Other members of the Cdk family, such as Cdk1 and Cdk2,were also capable of phosphorylating FOXO1 at Ser249 (Fig. 1B).By in vitro kinase assay, we found that the phosphor-S249-FOXO1 antibody recognized recombinant FOXO1 that wasphosphorylated by Cdk5/p35, Cdk5/p25, Cdk1/cyclin B, orCdk2/cyclin A in vitro but did not recognize unphosphorylatedFOXO1-S249A mutant (Fig. 1C,D). This particular experimentdemonstrates that activated Cdk5 is able to directly phosphory-late FOXO1. We next investigated the physical interactionsamong FOXO1, Cdk5, and p35. By coimmunoprecipitation inboth directions, FOXO1 was found to associate with both Cdk5and p35 either when all three proteins were coexpressed or when

protein pairs (FOXO1/Cdk5 and FOXO1/p35) were coexpressedseparately (Fig. 1E). These data suggested that both Cdk5 and p35physically associate with FOXO1.

FOXO1 has been reported to be phosphorylated by Cdk1 orCdk2 at the S249 site in cancer cells or cerebellar granule cells(Huang et al., 2006; Yuan et al., 2008); however, the potential roleof this pathway in cortical neurons remains completely un-known. Considering that the active Cdk5 is specifically found inthe nervous system especially in adult neurons, we thereforeasked whether the phosphorylation of FOXO1 by Cdk5 altered itstranscription activity in primary cortical neurons. In primarypostmitotic cortical neurons, active Cdk5, but not kinase-dead

Figure 4. Cdk5 promotes an AKT-independent nuclear export of FOXO1. A, B, Primary cortical neurons were transfected with Flag-FOXO1 together with GFP, GFP-p25/Cdk5, GFP-p35/Cdk5,GFP-p25/Cdk5-KD, or GFP-p35/Cdk5-KD on DIV 3. On DIV 7, the neurons were treated with vehicle control (DMSO) (A) or LY294002 (B) for 16 h. After treatments, the cells were fixed to measure thesubcellular localization of FOXO1 by Flag-immunofluorescent staining. Representative images are shown on the left: green represents GFP; red represents Flag; blue represents DAPI. Quantificationof subcellular localization of FOXO1 is shown on the right. C, D, Primary cortical neurons were transfected with Flag-FOXO1-AAA or Flag-FOXO1-AAA-S249A together with GFP, GFP-p25/Cdk5,GFP-p35/Cdk5, GFP-p25/Cdk5-KD, or GFP-p35/Cdk5-KD. On DIV 7, the cells were fixed to measure the subcellular localization of FOXO1 by Flag-immunofluorescent staining. Representative imagesare shown on the left: green represents GFP; red represents Flag; blue represents DAPI. Quantification of subcellular localization of FOXO1 is shown on the right.

2628 • J. Neurosci., February 11, 2015 • 35(6):2624 –2635 Zhou, Li et al. • Cdk5 Mediates FOXO1 Localization

Cdk5, dramatically reduced the 3xIRS luciferase expression(Fig. 1F). To confirm that the decreased FOXO1 transcriptionalactivity was mediated by Cdk5-directed phosphorylation, theFOXO1-S249A mutant was used to replace the wild-type. Asshown in Figure 1F, Cdk5/p25 had no effect on the expression ofthe 3xIRS-luciferase reporter gene when the mutant FOXO1-S249A was used. Active Cdk2/cyclinA also lowered 3xIRS activity,but active Cdk1/cyclin B increased it. Other classic Cdk/cyclin com-plexes (Cdk3/cyclin C, Cdk4/cyclin D, or Cdk6/cyclin D) had nosignificant effect on FOXO1 transcriptional activity (Fig. 1G).

Activation of Cdk5 accelerates FOXO1 nuclear exportInactivation of FOXO1 is commonly associated with its exportfrom the nucleus. We have documented that Cdk5 can serve as aFOXO1 kinase, yet AKT can also phosphorylate FOXO1, pro-moting its interaction with 14-3-3� proteins, its cytoplasmic se-questration, and its inhibition as a transcription factor(Burgering and Kops, 2002; van der Heide et al., 2004). We foundthat introduction of Cdk5/p25 significantly increased the associ-ation between FOXO1 and 14-3-3� compared with control vec-tor. However, Cdk5/p25 failed to increase the interactionbetween FOXO1-S249A and 14-3-3� proteins consistent with thehypothesis that phosphor-S249 is one key 14-3-3� interacting resi-due (Fig. 2A).

To determine the consequences of endogenous activation ofCdk5 on FOXO1, we exposed cultures of primary cortical neu-rons to H2O2, which is known to induce the cleavage of p35 to p25and activate Cdk5 (Fig. 2B). H2O2 treatments significantly in-creased the phosphorylation level of S249-FOXO1 while having

little effect on the total FOXO1 proteinlevel (Fig. 2B,C). H2O2 administrationalso dramatically induced the nuclear ex-port of FOXO1 (Fig. 2D,E). When Cdk5activity was inhibited by roscovitine (aCdk5 activity inhibitor), the endogenousFOXO1 (Fig. 2F) or the exogenous GFP-FOXO1 (Fig. 2G) relocated from cyto-plasm to nucleus within 12 h. Comparedwith wild-type-FOXO1, roscovitine treat-ments failed to induce the nuclear reloca-tion of phosphomimetic FOXO1-S249Dmutant (Fig. 2H, I).

H2O2 treatment and the resulting shiftin FOXO1 localization also significantlydecreased the expression of the 3xIRS-luciferase reporter (Fig. 2J). Importantly,this effect of H2O2 was abolished eitherwhen Cdk5 was silenced by shRNA or themutant FOXO1-S249A was used insteadof wild-type FOXO1 (Fig. 2L). Similar re-sults were obtained when �-amyloid wasapplied to the cultured neurons instead ofH2O2 (Fig. 2K,M). These findings dem-onstrate that the Ser249 phosphorylationby Cdk5 is vital both for the cytoplasmicretention of FOXO1 and for suppressingits activity as a transcription cofactor. Thefact that roscovitine treatment blocks thedownregulation of the 3xIRS-luciferasereporter by H2O2 or �-amyloid (Fig.2 J,K) further implies that it is Cdk5 activ-ity, not just the protein, which is the key tothis effect.

Cdk5 phosphorylation of FOXO1 at S249 is sufficient for itscytoplasmic retention and reverses nuclear localization ofFOXO1 resulting from AKT inhibitionNuclear accumulation of FOXO1 can also be achieved by inhib-iting the AKT pathway (Burgering and Kops, 2002; van der Heideet al., 2004), and we have replicated these findings in our system.Significantly, whereas the nuclear localizations of both wild-typeand mutant (S249A) FOXO1 were enhanced in the presence ofthe AKT inhibitor LY294002 (Fig. 3A,B,D,E), we found that thephosphomimetic FOXO1-S249D mutant retained significant cy-toplasmic localization, even in the presence of the AKT inhibitor(Fig. 3C,F). To address the same issue in reverse, we used anAKT-nonphosphorylatable FOXO1 mutant by converting thethree AKT target residues to alanine (FOXO1-AAA). On its own,this mutant lacks the ability to associate with 14-3-3� and thus isimported to the nucleus (Fig. 3G). Blocking the Cdk5 sites inaddition (FOXO1-AAA-S249A) had no additional effect on pro-tein localization (Fig. 3H). However, when the Cdk5 site wasaltered to its phosphomimetic form (FOXO1-AAA-S249D), theprotein lost its exclusively nuclear localization and could befound in both cytoplasm and nucleus (Fig. 3I). The levels ofexpression of the 3xIRS and BIM reporters were consistent withthe localization data. FOXO1-AAA-S249D had a reduced effecton the reporter constructs compared with the AKT-non-phosphorylatable FOXO1-AAA mutant (Fig. 3 J,K). To demon-strate the Cdk5 dependence of this phenomenon, we showed thatLY294002-induced nuclear accumulation of wild-type FOXO1was abolished by Cdk5/p35 or Cdk5/p25, but not by Cdk5-KD/

Figure 5. Cdk5 inhibits FOXO1 transcriptional activity in an AKT-independent way. Primary cortical neurons were transfectedwith the GFP-FOXO1-WT, GFP-FOXO1–S249A, GFP-FOXO1-AAA, or GFP-FOXO1-AAA-S249A plasmid together with indicated vec-tors in A, B, C or control vector in combination with the 3xIRS luciferase reporter system. With or without drug treatments, theneurons were subjected to luciferase assay. Data are mean � SEM; n � 5. *p � 0.05.

Zhou, Li et al. • Cdk5 Mediates FOXO1 Localization J. Neurosci., February 11, 2015 • 35(6):2624 –2635 • 2629

p35 or Cdk5-KD/p25 (Fig. 4A,B). Thesame results were obtained when theAKT-unresponsive mutant FOXO1-AAAwas used (Fig. 4C). Furthermore, whenthe Cdk5 phosphorylation site onFOXO1-AAA was removed (FOXO1-AAA-S249A), activation of Cdk5/p25 orCdk5/p35 failed to translocate the nuclearFOXO1 into cytoplasm (Fig. 4D).

Consistent with the localization data,the LY294002-induced upregulation ofthe FOXO1 transcriptional activity wasnot observed in the presence of Cdk5/p35or Cdk5/p25 but still evident when p35 orp25 was coexpressed with the kinase-deadform of Cdk5 (Fig. 5A, red columns).However, when FOXO1 (FOXO1-S249A)failed to be phosphorylated by Cdk5, AKTinhibition by LY294002 remained capableof activating FOXO1’s transcriptional ac-tivity, even in the presence of Cdk5/p35 orCdk5/p25 (Fig. 5A, green columns). Aspredicted from the above results, AKT in-hibition has relatively little effect on theCdk5 phosphomimetic form of FOXO1(FOXO1-S249D) (Fig. 5B). Overexpres-sion of Cdk5/p35 or Cdk5/p25 decreasedFOXO1’s transcriptional activity on wild-type FOXO1-AAA but had no effect onthe FOXO1-AAA-S249A mutant (Fig.5C). These results are consistent withthe decreased nuclear localization of theFOXO1-AAA due to the activation ofCdk5/p35 or Cdk5/p25 (Fig. 4C,D).

As mentioned above, Cdk1 and Cdk2can both phosphorylate FOXO1 at S249,and both kinases can regulate FOXO1transcriptional activity. In exploring thespecificity of the Cdk5 effect, we foundthat, when either Cdk1 or Cdk2 wastransfected with their activating cyclins,they were not able to downregulate theAKT-insensitive FOXO1-AAA mutant,even though the S249 residue was pres-ent (Fig. 5C).

Cell cycle, cell death, and the interplayof AKT, Cdk5, and FOXO1We have previously shown that, in theCdk5-deficient brain, there is a distinctloss of cell cycle suppression in maturingcortical neurons accompanied by massivecell death (Cicero and Herrup, 2005). Wewondered whether abnormal activationof FOXO1 might be involved either ofthese processes. We first asked whetherdeficiency of Cdk5 affects any aspects ofthe FOXO1 phenotype. In neurons derived from the E16.5 brainsof Cdk5�/� or Cdk5�/� mice, FOXO1 immunoreactivity wasfound predominately in the cytoplasm. In the Cdk5-deficientmouse brain (Cdk5�/�), however, we observed strong FOXO1staining in neuronal nuclei (Fig. 6A). Similar results were ob-tained in vitro: the endogenous FOXO1 or the GFP-tagged

FOXO1 was found in the nucleus of cultured Cdk5�/� neurons(Fig. 6B,C). Furthermore, the endogenous FOXO1 protein wasfound predominately in the nucleus in cells of the Cdk5-deficientmice (Cdk5�/�) or when Cdk5 was downregulated by siRNAs inneuronal cultures derived from wild-type mice (Fig. 6D–F).FOXO1 localized in cytoplasm is normally inactive (Gan et al.,

Figure 6. Nuclear accumulation of FOXO1 in Cdk5-deficient neurons leads to the activation of FOXO1. A, Representative images ofFOXO1 immunoreactivity in cortical regions of E16.5 Cdk5 �/�, Cdk5�/�, and Cdk5 �/� mice. B, C, Cdk5 �/� and Cdk5 �/� neuronswere isolated from E16.5 embryos obtained from crosses of Cdk5�/�mice. B, Subcellular localization of endogenous FOXO1 in Cdk5 �/�

and Cdk5 �/� neurons was evaluated by FOXO1 immunofluorescence staining in Tuj1-positive neurons. LY294002 treatments inducednuclear accumulation of FOXO1. C, Subcellular localization of exogenous FOXO1 was evaluated following transfection of GFP-FOXO1 inCdk5 �/� and Cdk5 �/� neurons. Representative images are shown on the left, and quantification of subcellular localization of FOXO1 isshown on the right. D, E, The cytoplasmic and nuclear extracts were isolated from cortices of E16.5 Cdk5 �/� and Cdk5 �/� mice. Lysateswere immunoblotted with the indicated antibodies. E, The relative level of FOXO1 in total/cytoplasm/nucleus was quantified. Data aremean�SEM;n�3.*p�0.05.F,LysatesfromprimarycorticalneuronsweretransfectedwithCdk5-specificsiRNA(01,02,03)andcontrolsiRNA (NC) on DIV 3. On DIV 7, the cytosolic and nuclear extracts were isolated and immunoblotted with the indicated antibodies. G,Cdk5 �/� and Cdk5 �/� neurons were transfected with the 3xIRS luciferase reporter system on DIV 3. On DIV 7, luciferase activities for3xIRS were measured and analyzed. Data are mean � SEM; n � 5. *p � 0.05. The protein level of Cdk5 in lysates was also provided. H,Primary cortical neurons (DIV 3) were transfected with plasmids or specific siRNA targeting Cdk5 as indicated. On DIV 7, luciferase activitiesfor 3xIRS were measured and analyzed. Data are mean � SEM; n � 5. *p � 0.05. The protein level of Cdk5 in lysates was also provided.

2630 • J. Neurosci., February 11, 2015 • 35(6):2624 –2635 Zhou, Li et al. • Cdk5 Mediates FOXO1 Localization

Figure 7. Cdk5 deficiency induces nuclear accumulation of FOXO1 by inactivating AKT. A, Primary cortical neurons were transfected with control U6 or shcdk5 plasmid together with GFP-FOXO1-WT, GFP-FOXO1–S249D, or GFP-FOXO1–S249A on DIV 3, then immunostained with Cdk5 antibody on DIV 7. The subcellular localizations of GFP-FOXO1-WT, GFP-FOXO1–S249D, or GFP-FOXO1–S249A were measured in neurons displaying low Cdk5 expression; arrows indicate shcdk5 effective neurons. Representative images are shown on the left: green represents GFP; red represents Cdk5;blue represents DAPI. Quantification is shown on the right. B, Lysates of 293T cells transfected with Flag-FOXO1 together with control U6 or shcdk5 plasmid were immunoprecipitated with the Flagantibody and immunoblotted with the FOXO1 and 14-3-3� antibodies. C, D, Primary cortical neurons transfected with the GFP-FOXO1-WT or GFP-FOXO1–S249A plasmid together with the controlU6 or shcdk5 plasmid plus either the 3xIRS reporter (C) or Bim reporter gene (D). Neurons were subjected to luciferase assays. Data are mean�SEM; n�5. *p�0.05. E, F, N2a cells were transfectedwith Cdk5/p25, Cdk5/p35, or control Flag vectors. Forty-eight hours after transfection, the cells were treated with DMSO or 50 �M LY294002 for another 12 h. The lysates were immunoblotted withthe indicated antibodies. F, The relative levels of pS319-FOXO1 and pS473-Akt1 were quantified. Data are mean � SEM; n � 3. G, H, N2a cells were transfected with shRNA targeted to either Cdk5or foxo1. Scrambled shRNA was used as a control. Forty-eight hours after transfection, the lysates were immunoblotted with the indicated antibodies. H, The relative level of pS319-FOXO1 wasquantified. Data are mean � SEM; n � 5. *p � 0.05. I, J, Brain lysates from the cortices of E16.5 Cdk5 �/� and Cdk5 �/� mice were prepared, and the levels of pS249-FOXO1 were detectedfollowing immunoprecipitation with FOXO1 antibody. The other proteins were measured by directly immunoblotting with the indicated antibodies. J, The relative levels of pS249-FOXO1,pS319-FOXO1, and pS473-Akt1 were quantified. Data are mean � SEM; n � 3. *p � 0.05.

Zhou, Li et al. • Cdk5 Mediates FOXO1 Localization J. Neurosci., February 11, 2015 • 35(6):2624 –2635 • 2631

2005); after relocation to the nucleus, its transcription activityincreases. This is equally true in our system. Depletion of endog-enous Cdk5 in neurons increased the expression of the 3xIRS-luciferase reporter gene, which suggests that endogenous Cdk5normally suppresses FOXO1-dependent transcription (Fig.6G,H).

We next investigated whether the dephosphorylation of S249on FOXO1 is involved in any of these events. It is worth notingthat the phosphomimetic FOXO1-S249D mutant retained signif-icant cytoplasmic localization in Cdk5-deficient neurons. Whenwe transfected the Cdk5-inactive FOXO1 mutant (S249A) intoCdk5-deficient neurons, it localized to the nucleus as did thewild-type FOXO1 (Fig. 7A); and knockdown Cdk5 also reducedthe interaction between FOXO1 with 14-3-3� (Fig. 7B). The ac-tivities of wild-type FOXO1 or FOXO1-S249A on the 3xIRS-luciferase reporter gene and a reporter gene controlled by the Bimpromoter were both increased by Cdk5 knockdown (Fig. 7C,D).These effects may be due in part to the inactivation of AKT underconditions of Cdk5 deficiency. First, increasing Cdk5 activity(overexpression of Cdk5/p35 or Cdk5/p25) did not affect theactivation of AKT as assessed by the phosphorylation of S473 andthe phosphorylation of FOXO1 on S319 (one of the phosphory-lation targets of AKT) (Fig. 7E,F). This suggests that active Cdk5promotes the cytoplasmic translocation of FOXO1 throughmechanisms independent of AKT regulation. However, whenCdk5 was silenced by shRNA, phosphorylation of FOXO1 at theAKT site (S319) was significantly decreased (Fig. 7G,H). Consis-tent with the in vitro observation, phosphorylation at S473 ofAKT and S319 of FOXO1 was significantly decreased in Cdk5�/�

brain lysates (Fig. 7 I, J). As expected, the phosphorylation at S249of FOXO1 was also decreased in Cdk5�/� brain lysates (Fig.7 I, J). Thus, it would appear that the nuclear accumulation ofFOXO1 under conditions of Cdk5 deficiency is in part mediatedby inhibition of AKT. These effects are summarized in Table 1. In-hibition of Cdk5, either genetically or pharmacologically, permitsthe nuclear accumulation of FOXO1. Activation of Cdk5 promotesFOXO1 cytoplasmic localization even if AKT is inhibited.

With respect to cell cycle regulation, we have previouslyshown that Cdk5 is vital for keeping postmitotic neurons fromreentering the cell cycle (Zhang and Herrup, 2008, 2011; Zhang etal., 2010b, 2012). To pursue these findings further, we culturedcortical neurons from Cdk5�/� mice for 7 DIV. As expected,spontaneous neuronal death and cell cycle reactivation were bothdetected. As shown in Figure 8A, Cdk5-deficient postmitoticneurons (TuJ1-positive) exhibit two distinct phenotypes: neuro-nal cell cycle reactivation (BrdU-positive) and neuronal death(activated caspase-3). We previously reported that Cdk5 nega-tively regulates E2F1, blocking neuronal cell cycle reentry; how-

ever, little is known regarding the molecular mechanismunderlying neuronal death phenotype in Cdk5�/� neurons. Toaddress this, we silenced foxo1 and Cdk5 in primary neurons andexamined effects on neuronal viability and BIM expression. Wefound that silencing of FOXO1 markedly diminished neuronalcell death under Cdk5 deficiency conditions by decreasing BIMexpression (Fig. 8B,D,E). To further interrogate this pathway, weperformed rescuing experiments by forced expression of varioussilencing-resistant forms of FOXO1. We found that Cdk5deficiency-induced apoptosis was restored by expression ofsilencing-resistant FOXO1 proteins WT-Res or S249A-Res, butnot cytoplasmic localized S249D-Res (Fig. 8C,D). Together, theseresults indicate that neuronal apoptosis induced by the loss ofCdk5 is indeed mediated by the nuclear localization of FOXO1.

Our previous data demonstrated that forcing a nuclear local-ization for Cdk5 (with an NLS) can block neuronal cell cyclereactivation but cannot stop apoptosis (Zhang et al., 2008, 2011).We find that cytoplasmic Cdk5 attenuates the neuronal death buthas no effect on the cell cycle reactivation in Cdk5�/� postmitoticneurons (Fig. 8F). This cytoplasmic effect proved to be tightlycorrelated with FOXO1 behavior. Further, increasing Cdk5 levelsin cytoplasm (Cdk5-NES) significantly decreased the FOXO1-dependent transcription of 3xIRS- and Bim reporters, whereasnuclear Cdk5 (Cdk5-NLS) was ineffective in this assay (Fig. 8G).

DiscussionCdk5 is an atypical cyclin-dependent kinase that acts like adouble-edged sword with respect to the regulation of neuronalsurvival. Under conditions such as oxidative stress or after chal-lenge with an increased �-amyloid burden in AD, enhanced levelsof calcium activate calpain. This drives the cleavage of p35 to p25and reportedly hyperactivates Cdk5, resulting in abnormal phos-phorylation of its substrates and eventually neuronal death (Pat-rick et al., 1999; Ahlijanian et al., 2000; Liu et al., 2003; Cruz andTsai, 2004). Yet other reports have demonstrated that Cdk5 ac-tivity is indeed vital for neuronal survival. Although the two ef-fects are not mutually exclusive, our laboratory and others havefound that Cdk5-deficient neurons are more vulnerable thanwild-type cells to DNA-damaging agents and to excitotoxic neu-ronal death (Cicero and Herrup, 2005; O’Hare et al., 2005;Turner et al., 2008; Zhang et al., 2008). Until now, the detailmechanism behind this vulnerability remained unclear. The datareported here provide fresh insights into the functions of bothCdk5 and FOXO1 in the survival of postmitotic neurons.

FOXO1 is normally maintained in the neuronal cytoplasmthrough its interaction with the 14-3-3� protein. This interactionrequires AKT-mediated FOXO1 phosphorylation on amino ac-ids T24, S256, and S319 (Arden, 2004). The involvement of Cdk5in the activation of AKT has been previously reported by Li et al.(2003) and Luo et al. (2014) who demonstrated that the phos-phorylation S473 on AKT (the activation site of AKT) was signif-icantly decreased in Cdk5-deficient mouse brains. Consistentwith their data, we find that suppression of Cdk5 downregulatesAKT activity (Fig. 7), which allows FOXO1 to dissociate from the14-3-3� protein, permitting FOXO1 to translocate to the nu-cleus. We postulate that AKT determines the localization ofFOXO1when Cdk5 activity is abolished. In the absence of Cdk5,FOXO1 is dephosphorylated and freed from its cytoplasmic lo-cation. Once it translocates to the nucleus, it triggers the transla-tion of Bim and hence drives the apoptotic process.

Nevertheless, Cdk5 can also regulate cytoplasmic localization ofFOXO1 through direct phosphorylation of FOXO1 at Ser249 inde-pendent of AKT (Fig. 4; Table 1). Our data thus emphasize the com-

Table 1. Summary of FOXO1 localization in primary cortical neuronsa

FOXO1 249A 249D AAA AAA-249A AAA-249D

Cdk5 �/� C C C N N C � NCdk5 �/� C � N C � N C N N C � NRoscovitine C � N C C N N C � NCdk5/p35 or Cdk5/p25 C C C C � N N C � NCdk5-KD/p25 or Cdk5-KD/p35 C C C N N C � NLY294002 N N C � N N N C � NCdk5/p35 or Cdk5/

p25 � LY294002C � N N C � N

Cdk5-KD/p35 or Cdk5-KD/p25 � LY294002

N

aThe indicated plasmids were transfected into Cdk5 �/� or Cdk5 �/� neurons on DIV 3; drug treatment wasadministered if desired on DIV 5. The localizations of FOXO1 were detected on DIV 7.

2632 • J. Neurosci., February 11, 2015 • 35(6):2624 –2635 Zhou, Li et al. • Cdk5 Mediates FOXO1 Localization

plex nature of the interaction between Cdk5 and AKT in controllingFOXO1 localization and hence its activity: Cdk5 phosphorylation ofS249 on FOXO1 is sufficient for its cytoplasmic retention, whereasthe phosphorylation of FOXO1 by AKT is necessary for its nuclearlocalization.

Cyclin-dependent kinases are the catalytic subunits of a familyof serine/threonine kinases that have been implicated in the con-trol of cell-cycle progression and neuronal function (Malumbres

and Barbacid, 2005). Cdk2 is reported tophosphorylate FOXO1 at Ser249 in tumorcells. This phosphorylation by Cdk2 re-sulted in FOXO1 cytoplasmic localizationand its inhibition (Huang et al., 2006).Cdk1, another classic Cdk, has also been re-ported to phosphorylate FOXO1 at Ser249when it is overexpressed in cerebellargranule cells. These conflicting resultssuggest caution in generalizing the effectsof Cdk kinases on FOXO1 activity. In ourmodel system, we have compared the effectof several Cdk/cyclin complexes on FOXO1.Overexpression of active Cdk1/cyclin Bdid increase FOXO1-dependent transcrip-tion, whereas Cdk2/cyclin A decreasedFOXO1 activity as Cdk5/p35 and Cdk5/p25 did (Fig. 1). The paradoxical resultsbetween cyclin Cdk1/cyclin B and Cdk5/p25 on IRS luciferase are unlikely due tocytotoxicity because we did not observeany significant difference in cell death be-tween cells overexpressing Cdk1/cyclin Bor Cdk5/p25 (data not shown). These dataemphasize that, even though all Cdkfamily members share the same phos-phorylation motif, their choice of spe-cific substrate is likely to be morecomplex. One theoretical advantage ofthis arrangement is that, given their po-tency in driving the cell cycle, Cdk1 andCdk2 activity must be maintained at verylow levels in cell cycle quiescent cells, suchas postmitotic neurons. By contrast,Cdk5, which is incapable of driving cellcycle activity, can stay high and realize itsfull benefit in the support of neuronalmaturation and survival. Indeed, as weand others have shown, its presence inthe nucleus actually serves to keep neu-rons out of the cell cycle (Jessberger etal., 2008, 2009; Zhang et al., 2008,2010a). From this perspective, it makessense for Cdk5 to assume the major roleas FOXO1 regulator in postmitotic neu-rons.

This study also contributes to our un-derstanding of two characteristics ofCdk5�/� postmitotic neurons: spontane-ous neuronal death and neuronal cell cy-cle reactivation. In adult neurons, cellcycle reactivation will induce neuronaldeath, but this cell cycle-related neuro-nal death is a slow progress and is dis-tinct from classic neuronal apoptosis(Fig. 8A) (Herrup and Yang, 2007). Our

new data clearly demonstrated that silencing of foxo1 byshRNA markedly diminished neuronal cell death (Fig. 8 B, D).More importantly, Cdk5 deficiency-induced apoptosis was re-stored by forced expression of silencing-resistant FOXO1 con-structs (FOXO1-WT-Res or FOXO1-S249A-Res), whereasFOXO1-S249D-Res failed to restore the neuronal apoptosiswhen both Cdk5 and FOXO1 were silenced. We showed that

Figure 8. Interaction of Cdk5-dependent cell cycle and cell death activities. A, Cleaved-caspase-3 and BrdU incorporation wasdouble stained on DIV 7 Cdk5 �/� neurons. TuJ1 staining serves as neuronal cell marker. B–D, Effects of expressing mutant orwild-type FOXO1 on cortical neuronal apoptosis. B, Primary cortical neurons (DIV 3) were transfected with CMV-mCherry-IRES-U6-shcdk5 or control CMV-mCherry-IRES-U6-scramble vector either alone or together with U6-shfoxo plasmid. Cells were then im-munostained with Cdk5 antibody at DIV 7. DAPI was used to counterstain nuclei to measure the neuronal death by nucleicondensation. C, Primary cortical neurons (DIV 3) were transfected with CMV-mCherry-IRES-U6-shcdk5 and shfoxo1 vector to-gether with GFP-FOXO1-WT-Res, GFP-FOXO1–S249A-Res, or GFP-FOXO1–S249D-Res plasmid. The cells were fixed on DIV 7 andstained with DAPI to evaluate neuronal death by nuclei condensation. D, The quantification of neuronal apoptotic rate in B, C ispresented in D. Data are mean � SEM; n � 3. *p � 0.05. F, Primary cortical Cdk5 �/� neurons were transfected with GFP-Cdk5-NLS or GFP-Cdk5-NES. Cleaved-caspase-3 and BrdU incorporation was detected by immunostaining. E, G, Primary cortical neuronswere transfected with the indicated plasmids together with the 3xIRS-luciferase or Bim-luciferase reporter gene. The neurons weresubjected to luciferase assays on DIV 7. Data are mean � SEM; n � 5. *p � 0.05.

Zhou, Li et al. • Cdk5 Mediates FOXO1 Localization J. Neurosci., February 11, 2015 • 35(6):2624 –2635 • 2633

FOXO1-S249D-Res retained in the cytoplasm, which limitedthe activation of BIM expression in Cdk5-deficient neurons,thus attenuating neuronal death (Fig. 8C,D).

The distribution of Cdk5 in nucleus and cytoplasm correlateswith the cell cycle reactivation and neuronal death, respectively.Extranuclear Cdk5 suppresses the cell cycle reactivation, but itdoes not attenuate the quick neuronal death (Zhang and Herrup,2011; Zhang et al., 2010a, b). As shown in this current work,cytoplasmic Cdk5 blocks apoptosis by inactivating FOXO1 todecrease BIM levels but has no affect on the cell cycle reentry inCdk5�/� neurons (Fig. 8G).

In conclusion, our data demonstrate that Cdk5 is a key to thecytoplasmic maintenance of FOXO1, keeping it inactive and thusprotecting postmitotic cortical neurons from neuronal death.These findings offer important new insights into the roles thatCdk5 plays in the adult CNS and have implications for the etiol-ogy of neurodegenerative diseases, such as AD.

ReferencesAhlijanian MK, Barrezueta NX, Williams RD, Jakowski A, Kowsz KP, McCar-

thy S, Coskran T, Carlo A, Seymour PA, Burkhardt JE, Nelson RB, Mc-Neish JD (2000) Hyperphosphorylated tau and neurofilament andcytoskeletal disruptions in mice overexpressing human p25, an activatorof cdk5. Proc Natl Acad Sci U S A 97:2910 –2915. CrossRef Medline

Arden KC (2004) FoxO: linking new signaling pathways. Mol Cell 14:416 –418. CrossRef Medline

Burgering BM, Kops GJ (2002) Cell cycle and death control: long live Fork-heads. Trends Biochem Sci 27:352–360. CrossRef Medline

Cheung ZH, Ip NY (2004) Cdk5: mediator of neuronal death and survival.Neurosci Lett 361:47–51. CrossRef Medline

Cicero S, Herrup K (2005) Cyclin-dependent kinase 5 is essential for neu-ronal cell cycle arrest and differentiation. J Neurosci 25:9658 –9668.CrossRef Medline

Cruz JC, Tsai LH (2004) A Jekyll and Hyde kinase: roles for Cdk5 in braindevelopment and disease. Curr Opin Neurobiol 14:390 –394. CrossRefMedline

Dhavan R, Tsai LH (2001) A decade of CDK5. Nat Rev Mol Cell Biol 2:749 –759. CrossRef Medline

Essafi A, Fernandez de Mattos S, Hassen YA, Soeiro I, Mufti GJ, Thomas NS,Medema RH, Lam EW (2005) Direct transcriptional regulation of Bimby FoxO3a mediates STI571-induced apoptosis in Bcr-Abl-expressingcells. Oncogene 24:2317–2329. CrossRef Medline

Gan L, Zheng W, Chabot JG, Unterman TG, Quirion R (2005) Nuclear/cytoplasmic shuttling of the transcription factor FoxO1 is regulated byneurotrophic factors. J Neurochem 93:1209 –1219. CrossRef Medline

Gilmore EC, Herrup K (2001) Neocortical cell migration: GABAergic neu-rons and cells in layers I and VI move in a cyclin-dependent kinase5-independent manner. J Neurosci 21:9690 –9700. Medline

Hawasli AH, Bibb JA (2007) Alternative roles for Cdk5 in learning and syn-aptic plasticity. Biotechnol J 2:941–948. CrossRef Medline

Hellmich MR, Pant HC, Wada E, Battey JF (1992) Neuronal cdc2-like ki-nase: a cdc2-related protein kinase with predominantly neuronal expres-sion. Proc Natl Acad Sci U S A 89:10867–10871. CrossRef Medline

Herrup K, Yang Y (2007) Cell cycle regulation in the postmitotic neuron:oxymoron or new biology? Nat Rev Neurosci 8:368 –378. CrossRefMedline

Hisanaga S, Endo R (2010) Regulation and role of cyclin-dependent kinaseactivity in neuronal survival and death. J Neurochem 115:1309 –1321.CrossRef Medline

Huang H, Regan KM, Lou Z, Chen J, Tindall DJ (2006) CDK2-dependentphosphorylation of FOXO1 as an apoptotic response to DNA damage.Science 314:294 –297. CrossRef Medline

Jessberger S, Aigner S, Clemenson GD Jr, Toni N, Lie DC, Karalay O, OverallR, Kempermann G, Gage FH (2008) Cdk5 regulates accurate matura-tion of newborn granule cells in the adult hippocampus. PLoS Biol 6:e272.CrossRef Medline

Jessberger S, Gage FH, Eisch AJ, Lagace DC (2009) Making a neuron: Cdk5in embryonic and adult neurogenesis. Trends Neurosci 32:575–582.CrossRef Medline

Lehtinen MK, Yuan Z, Boag PR, Yang Y, Villen J, Becker EB, DiBacco S, de la

Iglesia N, Gygi S, Blackwell TK, Bonni A (2006) A conserved MST-FOXO signaling pathway mediates oxidative-stress responses and extendslife span. Cell 125:987–1001. CrossRef Medline

Lew J, Huang QQ, Qi Z, Winkfein RJ, Aebersold R, Hunt T, Wang JH (1994)A brain-specific activator of cyclin-dependent kinase 5. Nature 371:423–426. CrossRef Medline

Li BS, Zhang L, Takahashi S, Ma W, Jaffe H, Kulkarni AB, Pant HC (2002)Cyclin-dependent kinase 5 prevents neuronal apoptosis by negative reg-ulation of c-Jun N-terminal kinase 3. EMBO J 21:324 –333. CrossRefMedline

Li BS, Ma W, Jaffe H, Zheng Y, Takahashi S, Zhang L, Kulkarni AB, Pant HC(2003) Cyclin-dependent kinase-5 is involved in neuregulin-dependentactivation of phosphatidylinositol 3-kinase and Akt activity mediatingneuronal survival. J Biol Chem 278:35702–35709. CrossRef Medline

Liu F, Su Y, Li B, Zhou Y, Ryder J, Gonzalez-DeWhitt P, May PC, Ni B (2003)Regulation of amyloid precursor protein (APP) phosphorylation andprocessing by p35/Cdk5 and p25/Cdk5. FEBS Lett 547:193–196. CrossRefMedline

Luo F, Burke K, Kantor C, Miller RH, Yang Y (2014) Cyclin-dependentkinase 5 mediates adult OPC maturation and myelin repair through mod-ulation of Akt and GsK-3beta signaling. J Neurosci 34:10415–10429.CrossRef Medline

Malumbres M, Barbacid M (2005) Mammalian cyclin-dependent kinases.Trends Biochem Sci 30:630 – 641. CrossRef Medline

Modur V, Nagarajan R, Evers BM, Milbrandt J (2002) FOXO proteins reg-ulate tumor necrosis factor-related apoptosis inducing ligand expression:implications for PTEN mutation in prostate cancer. J Biol Chem 277:47928 – 47937. CrossRef Medline

O’Hare MJ, Kushwaha N, Zhang Y, Aleyasin H, Callaghan SM, Slack RS,Albert PR, Vincent I, Park DS (2005) Differential roles of nuclear andcytoplasmic cyclin-dependent kinase 5 in apoptotic and excitotoxic neu-ronal death. J Neurosci 25:8954 – 8966. CrossRef Medline

Ohshima T, Ward JM, Huh CG, Longenecker G, Veeranna, Pant HC, BradyRO, Martin LJ, Kulkarni AB (1996) Targeted disruption of the cyclin-dependent kinase 5 gene results in abnormal corticogenesis, neuronalpathology and perinatal death. Proc Natl Acad Sci U S A 93:11173–11178.CrossRef Medline

Ohshima T, Gilmore EC, Longenecker G, Jacobowitz DM, Brady RO, HerrupK, Kulkarni AB (1999) Migration defects of cdk5(�/�) neurons in thedeveloping cerebellum is cell autonomous. J Neurosci 19:6017– 6026.Medline

Patrick GN, Zukerberg L, Nikolic M, de la Monte S, Dikkes P, Tsai LH (1999)Conversion of p35 to p25 deregulates Cdk5 activity and promotes neuro-degeneration. Nature 402:615– 622. CrossRef Medline

Real PJ, Benito A, Cuevas J, Berciano MT, de Juan A, Coffer P, Gomez-Roman J,Lafarga M, Lopez-Vega JM, Fernandez-Luna JL (2005) Blockade of epider-mal growth factor receptors chemosensitizes breast cancer cells through up-regulation of Bnip3L. Cancer Res 65:8151–8157. CrossRef Medline

Su SC, Tsai LH (2011) Cyclin-dependent kinases in brain development anddisease. Annu Rev Cell Dev Biol 27:465– 491. CrossRef Medline

Takahashi S, Ohshima T, Hirasawa M, Pareek TK, Bugge TH, Morozov A,Fujieda K, Brady RO, Kulkarni AB (2010) Conditional deletion of neu-ronal cyclin-dependent kinase 5 in developing forebrain results in micro-glial activation and neurodegeneration. Am J Pathol 176:320 –329.CrossRef Medline

Tanaka T, Veeranna, Ohshima T, Rajan P, Amin ND, Cho A, Sreenath T, PantHC, Brady RO, Kulkarni AB (2001) Neuronal cyclin-dependent kinase5 activity is critical for survival. J Neurosci 21:550 –558. Medline

Tsai LH, Delalle I, Caviness VS Jr, Chae T, Harlow E (1994) p35 is a neural-specific regulatory subunit of cyclin-dependent kinase 5. Nature 371:419 – 423. CrossRef Medline

Turner NC, Lord CJ, Iorns E, Brough R, Swift S, Elliott R, Rayter S, Tutt AN,Ashworth A (2008) A synthetic lethal siRNA screen identifying genesmediating sensitivity to a PARP inhibitor. EMBO J 27:1368 –1377.CrossRef Medline

van der Heide LP, Hoekman MF, Smidt MP (2004) The ins and outs ofFoxO shuttling: mechanisms of FoxO translocation and transcriptionalregulation. Biochem J 380:297–309. CrossRef Medline

Xie Q, Hao Y, Tao L, Peng S, Rao C, Chen H, You H, Dong MQ, Yuan Z(2012) Lysine methylation of FOXO3 regulates oxidative stress-inducedneuronal cell death. EMBO Rep 13:371–377. CrossRef Medline

Yuan Z, Becker EB, Merlo P, Yamada T, DiBacco S, Konishi Y, Schaefer EM,

2634 • J. Neurosci., February 11, 2015 • 35(6):2624 –2635 Zhou, Li et al. • Cdk5 Mediates FOXO1 Localization

Bonni A (2008) Activation of FOXO1 by Cdk1 in cycling cells and post-mitotic neurons. Science 319:1665–1668. CrossRef Medline

Yuan Z, Lehtinen MK, Merlo P, Villen J, Gygi S, Bonni A (2009) Regulationof neuronal cell death by MST1-FOXO1 signaling. J Biol Chem 284:11285–11292. CrossRef Medline

Zhang J, Herrup K (2008) Cdk5 and the non-catalytic arrest of the neuronalcell cycle. Cell Cycle 7:3487–3490. CrossRef Medline

Zhang J, Herrup K (2011) Nucleocytoplasmic Cdk5 is involved in neuronalcell cycle and death in post-mitotic neurons. Cell Cycle 10:1208 –1214.CrossRef Medline

Zhang J, Cicero SA, Wang L, Romito-Digiacomo RR, Yang Y, Herrup K(2008) Nuclear localization of Cdk5 is a key determinant in the postmi-

totic state of neurons. Proc Natl Acad Sci U S A 105:8772– 8777. CrossRefMedline

Zhang J, Li H, Herrup K (2010a) Cdk5 nuclear localization is p27-dependent in nerve cells: implications for cell cycle suppression andcaspase-3 activation. J Biol Chem 285:14052–14061. CrossRef Medline

Zhang J, Li H, Yabut O, Fitzpatrick H, D’Arcangelo G, Herrup K (2010b)Cdk5 suppresses the neuronal cell cycle by disrupting the E2F1-DP1 com-plex. J Neurosci 30:5219 –5228. CrossRef Medline

Zhang J, Li H, Zhou T, Zhou J, Herrup K (2012) Cdk5 levels oscillate duringthe neuronal cell cycle: Cdh1 ubiquitination triggers proteosome-dependent degradation during S-phase. J Biol Chem 287:25985–25994.CrossRef Medline

Zhou, Li et al. • Cdk5 Mediates FOXO1 Localization J. Neurosci., February 11, 2015 • 35(6):2624 –2635 • 2635